1. The CT-Based Fractal Analysis Of Trabecular Bone Structure May Help In Detecting Decreased Quality Of Bone Prior To Urgent Spinal Procedures
Czyz M 1, Kapinas A 2, Holton J 1 Pyzik R 3, Boszczyk B 2, Quraishi NA 2.
Spinal Service, The Royal Orthopaedic Hospital NHS Trust, Birmingham, UK.
The Centre for Spinal Studies and Surgery, Nottingham University Hospitals NHS Trust, Nottingham, UK.
Department of Radiology, Icahn School of Medicine at Mount Sinai, New York, USA.
Commonly accepted methods of accessing the bone density are quantitative computerised tomography (qCT) and dual-energy xray absorptiometry (DEXA) with qCT being the more accurate method (7)
pull out strength for those with the greater bone mineral density as assessed by qCT and DEXA (8).
In patients who have poor bone mineral density and require instrumentation there is the option of cement augmentation that has been shown to increase the pull out strength of screws (9). Technical aspects of surgery can also be altered in patients with known osteoporosis, such as reduced tapping of the screw prior to insertion (6). Larger and longer screws may also increase hold in osteoporotic bone (9).
This pre-operative planning may not be available for emergency procedures and therefore there is a requirement for a cheap and readily available method of assessment of the bone structure for planning when formal pre-operative assessments of bone mineral density cannot be achieved.
Fractal analysis is a method of quantifying a repeating pattern or structure and can be applied to biological patterns such as trabecular bone (10).
The analysis provides an assessment of the microarchitecture and trabecular structure of cancellous bone thus providing a direct assessment of the degree of osteoporosis (11,12). Cadaveric studies have shown that fractal analysis is a sensitive tool to identify osteoporosis (13). Dougherty et al has extended this into clinical studies and has shown that axial images with CT or MRI are sufficient to detect differences in architectural properties of trabecular bone and hence the degree of osteoporosis by fractal analysis (14).
Our aim is to use fractal analysis of axial CT images of the lumbar vertebrae as a quick and readily available method of quantifying the bone mineral density prior to urgent or semi-urgent spinal procedures.
Only patients who had a dual-energy x-ray absorptiometry (DEXA) scan performed within six months prior of after the CT-scan were subsequently enrolled into the study.
Axial scans of the lumbar spine were extracted from the PACS and stored in stack of TIFF images. Basic demographic data, bone mineral density (BMD) and T-score (number of standard deviations the BMD deviates from the bone mass of an average healthy 30 year old adult) were recorded for all individuals in the cohort. Diagnosis of osteopenia (OPA) or osteoporosis (OST) was made if the average T-score of the lumbar spine was lower than -1.0 and -2.5 respectively.
thresholding and binarization were performed. To ensure that the treatment was standardized, thresholding was achieved using the standard ‘IsoData’ protocol [3]. This algorithm allowed to achieve a clear selection between the bone (cancellous and cortical) and bone marrow as well as surrounding soft tissues. Subsequently, the area outside of the ROI assigned to a particular scan was deleted.
Fractal analysis of the segmented elements of the vertebra was performed using the box-counting method available in the FracLac plug-in (Karperien, A., version 2.5, 1e) [4]. A grid of known scale was automatically placed over the image and the number of boxes that contained non-zero pixels was counted (border pixels = 1, background pixels = 0).
This process was then repeated for multiple grids with increasing spacing. As the scale increases, the number of boxes containing the object decreases exponentially and the exponent is equivalent to the FD. To quantify the exponent, natural logarithmic plots of the number of boxes against scale and the gradient (−FD) were estimated using linear regression.
Fd is a number 1-2 to describe a complex image- solid white = 2 solid line 1
Tree fd close to 1- higher you go small branches higher fd goes
Introduction
To date no reliable method is available to determine the parameters of bone density based on the routine spinal CT in the emergency setup. We propose the use of fractal analysis to detect patients with poor quality of bone prior to urgent or semi-urgent spinal procedures.
Purpose:
To validate the hypothesis that the CT-based fractal analysis of the trabecular bone structure may help in detecting patients with poor quality of bone prior to urgent spinal procedures.
Study Design:
Retrospective analysis of prospectively collected data.
Figure 2: Plots illustrating distribution of median values of the fractal dimension (FD: left-sided graph) among vertebrae of normal and decreased quality of bone (T-score < -1.0). On the right side median fractal dimension (mFD) among groups of patients found with normal and decreased quality of bone (mT-score < -1.0). The central box represents the values from the lower to upper quartile (25 to 75 percentile). The middle line represents the median.
A line extends from the minimum to the maximum value, excluding “outside” and “far out” values which are displayed as separate points.
Methods
Patients in whom the DEXA scan and lumbar spine CT were performed at an interval of no more than three months were randomly selected from a prospectively collected database.
Diagnostic axial CT scans of L2, L3 and L4 vertebrae were processed to determine the fractal dimension (FD) of the trabecular structure of each spinal level.
Box-count method and ImageJ 1.49 software were used
The FD was compared to the results of the DEXA scan: bone mineral density (BMD) and T-score by mean of correlation coefficients. ROC curve analysis was later performed in order to determine cut-of value of FD.
Data analysis was performed using Medcalc® v14.8
Figure 3: Receiver operating characteristics (ROC) curves for fractal dimension (FD) indicating T-score < -1.0 (left), median fractal dimension (mFD) indicating median T-score < -1.0 (middle) and mFD indicating presence of at least one vertebra with T-score < -1.0 (right).
The cutoff values were estimated at 1.53 in FD group and 1.46 in mFD groups.
Results
A total of 102 vertebral levels obtained from 35 patients (mean age 60 ± 18 years; 29 female) were analysed.
The FD was significantly higher in the group of patients with decreased bone density [DBD] (T-score < -1.0) (1.67 vs. 1.43; P < ) and negatively correlated with BMD (R-Spearman -0.53; P < ) and T-score (-0.49; P < ).
ROC curve analysis revealed cut-off value of FD > 1.53 indicating DBD (P < ; AUC 0.84; 95%CI, ).
Conclusions
This study shows that fractal analysis of the lumbar spine CT images may be used to determine bone density prior to spinal instrumentation (e.g. metastatic/traumatic cord compression).
Further prospective studies are warranted comparing results of the fractal analysis of CT scans with qCT.
Figure 1: Image processing sequence and fractal analysis of a vertebral body. Example analysis of a single scan (1) out of a CT stack, belonging to a patient with osteoporosis. Automatic thresholding and binarization
(2) is followed by fractal analysis. The reflection of cortex and trabecular structure is being divided into boxes of increasingly smaller size (3-5), the number of boxes of size ε required to cover the edge was counted and indicated as N(ε). Fractal dimension (FD) is given by: FD = limε→0[log N(ε)/log(ε)].
The authors report no conflicts of interest and no funding sources for this research. This study was carried out in full compliance with UK law.